|Publication number||US6914820 B1|
|Application number||US 10/956,906|
|Publication date||Jul 5, 2005|
|Filing date||Sep 30, 2004|
|Priority date||May 6, 2002|
|Publication number||10956906, 956906, US 6914820 B1, US 6914820B1, US-B1-6914820, US6914820 B1, US6914820B1|
|Inventors||Sau Ching Wong|
|Original Assignee||Multi Level Memory Technology|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (85), Referenced by (15), Classifications (31), Legal Events (5)|
|External Links: USPTO, USPTO Assignment, Espacenet|
This patent document is a continuation of U.S. patent application Ser. No. 10/831,675, filed Apr. 23, 2004 now U.S. Pat. No. 6,826,084, which is a divisional of U.S. patent application Ser. No. 10/140,527, filed May 6, 2002, now U.S. Pat. No. 6,747,896. This patent document claims benefit of the earlier filing dates of and hereby incorporates by reference in their entirety both of the related applications.
One of the primary goals of memory manufacturers is increasing the storage density of memory devices. Improvements in integrated circuit fabrication techniques can achieve this goal by reducing the sizes of integrated circuit structures. Accordingly, as fabrication techniques improve, manufacturers can often increase memory densities simply by making the same memory structures smaller. Another technique for improving storage density is improving the functionality of memory structures to provide more storage per area. This can be achieved, for example, by creating memory cells and peripheral memory circuits that are capable of storing more information per memory cell.
U.S. Pat. No. 6,011,725, entitled “Two Bit Non-Volatile Electrically Erasable and Programmable Semiconductor Memory Cell Utilizing Asymmetrical Charge Trapping” describes a non-volatile memory that stores two bits per memory cell.
Two bits of data are stored in memory cell 100 as charge that is trapped in separated and isolated locations 140A and 140B in nitride region 140. Each location 140A or 140B corresponds to a bit having a value 0 or 1 according to the state of trapped charge at the location 140A or 140B. To program cell 100, gate 150 is raised to a high voltage while a channel current passes between diffused regions 120A and 120B and injects charge into nitride region 140. The location 140A or 140B of the injected charge depends on the characteristics of memory cell 100, the applied voltages, and whether the channel current flows from region 120A to region 120B or from region 120B to region 120A. The direction of the channel current during a programming operation thus selects which of the bits (i.e., location 140A or 140B) is programmed.
Reading a data bit from a particular location 140A or 140B is accomplished by biasing gate 150 at a voltage that is above the threshold voltage of memory cell 100 when locations 140A and 140B are in an unprogrammed state. The diffused region 120A or 120B that is closest to the location 140A or 140B being read is biased as the source/region for the read operation. Any charge trapped in locations 140A and 140B affects a portion of the underlying channel so that negative charge trapped near the source effectively reduces the gate-to-source voltage and correspondingly reduces the channel current during the read operation. In contrast, negative charge near the drain region is ineffective at reducing the channel current since an appropriate drain voltage effectively punches through the portion of the channel near the drain. Sensing whether a channel current flows in memory cell 100 during the read indicates the value of the bit associated with the location 140A or 140B nearest the source/region 120A or 120B.
Memory cell 100 has the advantage of providing non-volatile storage of two bits of information in a single-transistor memory cell, increasing the storage density when compared to a memory device storing one bit of data per storage transistor. However, scaling memory cell 100 down to smaller feature sizes may present difficulties. In particular, operation of memory cell 100 requires the ability to inject charge into separate locations 140A and 140B in nitride region 140. As the size of nitride region 140 decreases, the shorter distance between locations 140A and 140B may be unable to accommodate lateral charge movement after the write operation. Additionally, the amount of charge trapped at locations 140A and 140B of nitride region 140 is relatively small (e.g., typically a few hundred electrons) when compared, for example, to the charge (e.g., typically tens of thousands of electrons) in the floating gate of a conventional Flash memory cell. The smaller trapped charge makes precise control of threshold voltages more difficult because small variations in the trapped charge have large effects. This renders analog or multi-bit storage at each location 140A or 140B in memory cell 100 substantially more difficult than analog or multi-bit storage in a conventional Flash memory cell.
In accordance with an aspect of the invention, a memory transistor has two laterally separated floating gates over a channel. A control gate that overlies the floating gates extends into a gap between the floating gates to directly modulate a central channel portion between the floating gates. The memory transistor can store separate data values as charge on the separate floating gates. The threshold voltage of the memory transistor depends on the charge stored on the floating gates and the direction of the channel current. Since the amount of charge that can be stored on each floating gate is relatively large compared to charge that can be trapped in a gate insulator, the amounts of stored charge and the threshold voltages of the dual-floating-gate memory transistor can be controlled more precisely than is possible in some known memory devices that store data as locally trapped charge. The control gate directly modulating the central channel region shuts off the current through unselected memory transistors, which permits “over-erasing” the floating gates to extend the usable threshold voltage range for storing data. The improved control of the threshold voltage and the larger available threshold voltage range facilitates reliable storage of multiple levels or multiple bits of data in each floating gate.
In accordance with a further aspect of the invention, the memory transistor having laterally separated floating gates uses holes in the floating gates to define the charge states representing data values. Charge states arising from holes on a floating gate are known to provide better data stability. The holes cause channel regions under the floating gate to have low or negative threshold voltages, while the central channel region, which the control gate modules, has a positive threshold voltage. Accordingly, the memory transistor is off when the control gate is grounded, but a read operation that biases the control gate to a level sufficient for charge inversion in the central channel region can compare the amount of current through a memory transistor to a reference current to determine a stored data value.
One specific embodiment of the invention is a device containing an array of memory transistors. Each memory transistor includes: a first source/drain region, a second source/drain region, and a channel in a substrate; a first floating gate overlying a first end of the channel adjacent the first source/drain region; a second floating gate overlying a second end of the channel adjacent the second source/drain region; and a control gate overlying the first and second floating gates and extending into the gap between the first and second floating gates. The first and second source/drain regions can extend under part of the first and second floating gates, respectively, to reduce the effective channel lengths under the first and second floating gates and improve the selectivity and precision of writing and reading stored data values associated with the floating gates.
In contactless, virtual ground architecture, the array includes multiple banks. Each bank includes diffused lines in the substrate, and each column of the memory transistors in the bank corresponds to and connects to an adjacent pair of the diffused lines. A first of the corresponding diffused lines electrically connects the first source/drain regions of the memory transistor in the row, and a second of the corresponding diffused lines electrically connects the second source/drain regions of the memory transistor in the row. Word lines overlie and connect to or form the control gates for the memory transistors in corresponding rows of the array.
Metal column lines overlie the banks and connect to the diffused lines through bank select devices. In particular, first bank select cells connect to respective column lines, and each first bank select cell is between the connected column line and a corresponding adjacent pair of the diffused lines. Second bank select cells also connect to the column lines with each second bank select cell being between the connected column line and a corresponding adjacent pair of the diffused lines. The first and second bank select cells connect to opposite ends of the diffused lines in the bank, and the adjacent pairs of diffused lines corresponding to the second bank select cells are offset relative to the adjacent pairs of diffused lines corresponding to the first bank select cells. With this configuration, the numbers of the column lines, the diffused lines, and the floating gates are in respective proportions N, 2N−1, and 4(N-1). The metal column lines, which connect to peripheral circuits, have a pitch that is wide compared to the pitch of metal lines in a conventional contactless Flash memory. The wider pitch provides additional area for layout of pitch-sensitive array supporting circuits and reduces capacitive coupling between metal column lines.
Another embodiment of the invention is an erase operation for a memory transistor having the above-described structure. The erase operation includes biasing the control gate and a well containing the memory transistor at respective negative and positive voltages that are sufficient to induce charge tunneling between the well and the first and second floating gates. The biasing of the control gate and the well is maintained to remove any excess electrons from the first and second floating gates and can be continued to over-erase the first and second floating gates. As a result, the first and second floating gates can have an excess of holes that gives the underlying channel regions negative threshold voltages and/or operation in depletion mode. The lower threshold voltage of the erased states for the memory transistors provides a wider threshold voltage range for analog or multi-bit data storage. High threshold voltages are not needed for data storage, which improves data retention, reduces cell disturb, and may avoid the need for word line boost circuits that can slow the biasing of word lines during random-access read operations.
Another embodiment of the invention is a write operation for a memory transistor such as described above. The write operation includes biasing the control gate, the first source/drain region, and the second source/drain region at a first programming voltage, ground, and a second programming voltage, respectively. The first and second programming voltages respectively on the control gate and the second source/drain region induce channel hot electron injection that injects electrons into the second floating gate without changing the charge on the first floating gate. The write operation can further include biasing the control gate, the first source/drain region, and the second source/drain region at the first programming voltage, the second programming voltage, and ground. The first and second programming voltages respectively on the control gate and the first source/drain region induce channel hot electron injection that injects electrons into the first floating gate without changing the charge on the second floating gate. The write operation can store a binary, analog, or multi-bit value on a floating gate by stopping the write operation when the floating gate reaches a charge state representing the value to be stored.
A series of verify operations can test whether a write operation has reached a target state corresponding to the value being stored. One verify operation biases the control gate at a first read voltage, grounds the second source/drain region, biases the first source/drain region at a second read voltage; and compares current through the memory transistor to a reference current associated with the multi-bit value. The first read voltage, which is applied to the control gate, is typically higher than the upper boundary of the threshold voltage range used to store data, which causes the memory transistor to be conductive regardless of the charge states of the floating gates. The write operation ends in response to the comparison indicating that the current through the memory transistor corresponds to a level associated with the value being written. An alternative verify operation can bias the control gate at the target threshold voltage for the memory transistor and then sense whether the memory transistor conducts.
Yet another embodiment of the invention is a read operation for a memory transistor having the structure described above. To read a data value associated with the first floating gate, the read operation includes: biasing the control gate at a first read voltage (typically higher than the highest threshold voltage used for data storage); grounding the first source/drain region; biasing the second source/drain region at a second read voltage; comparing a channel current of the memory transistor to one or more reference currents associated with stored values; and using results of the comparisons to determine a first stored value, which is associated with the first floating gate. To read a data value associated with the second floating gate, the read operation includes: biasing the control gate at the first voltage; grounding the second source/drain region; biasing the first source/drain region at the second voltage; comparing the channel current of the memory transistor to the one or more reference currents; and using results of these comparisons to determine a second stored value, which is associated with the second floating gate. The one or more reference currents can be a single reference current for storage of one bit or analog value per floating gate or multiple reference currents respectively corresponding to multi-bit stored values.
Yet another embodiment of the invention is a method for manufacturing a memory device. The method includes: forming a first source/drain region, a second source/drain/region, and a channel in a substrate, wherein the channel extends from the first source/drain region to the second source/drain region; forming a first floating gate overlying and insulated from a first portion of the channel adjacent the first source/drain region; forming a second floating gate overlying and insulated from a second portion of the channel adjacent the second source/drain region, wherein a gap between the second floating gate and the first floating gate overlies a central portion of the channel between the first and second portions of the channel; and forming a control gate overlying and insulated from the first and second floating gates, the control gate extending into the gap between the first and second floating gates and modulating the central portion of the channel.
The first and second source/drain regions can be formed before the first and second floating gates so that the first and second source/drain regions underlie significant portions of the first and second floating gates. Alternatively, the first and second source/drain regions can be formed by implanting impurities into the substrate using the first and second floating gates to at least partially define boundaries of implanted areas and then oxidizing the implanted regions at high temperature to cause the implanted regions to diffuse laterally under the first and second floating gates and to form oxide regions over the first and second source/drain regions. The first and second floating gates can also control implantation steps that adjust a threshold voltage of the central region relative to threshold voltages of the first and second portions of the channel.
Use of the same reference symbols in different figures indicates similar or identical items.
In accordance with an aspect of the invention, a memory transistor for non-volatile storage of multiple data bits has two floating gates laterally separated over a common channel. The memory transistor is bi-directional in that the threshold voltage of the memory transistor depends on the direction of the current through the channel and the charge state of the particular floating gate nearest the region acting as the source of channel current. To store data in a selected one of the two floating gates, a programming operation drives the channel current in a direction that injects charge into the selected floating gate. A read operation selects a direction for the channel current according to which of the two floating gates is being read.
In accordance with another aspect of the invention, extending the source/drain regions under the floating gates reduces the effective size of the floating gates and the effective channel length that the floating gates influence. The source/drain regions can be formed under the floating gates by forming the source/drain regions before forming the floating gates, implanting the source/drain regions through portions of the floating gates followed by a short high-temperature oxidation cycle, or heating the structure so that impurities from implanted regions diffuse laterally under the floating gates. The two floating gates thus overlie and affect small portions of the channel of the memory transistor. Mini-field oxide regions grown over the source/drain regions can separate portions of the floating gates from the source/drain regions to reduce capacitve couplings between the source/drain regions and the overlying floating gates and between the source/drain regions and the overlying control gate.
A photolithographic patterning and chemical etching process can define widths GA and GB of floating gates 240A and 240B and a separation GC between floating gates 240A and 240B, making widths GA and GB and separation GC at least as large as the minimum feature size of patterning process. In memory transistor 200′, edges of source/drain regions 220A and 220B are self-aligned with edges of floating gates 240A and 240B, so that channel regions CA′ and CB′ have lengths about equal to the respective widths GA and GB of the overlying floating gates 240A and 240B.
Matching performance of both bits in memory transistor 200′ generally requires that the widths GA and GB of floating gates 240A and 240B be equal and the lengths of channel regions CA′ and CB′ be equal. The geometry of memory transistor can however be widely varied from the geometry of the illustrated embodiment. For example, separation GC between floating gates 240A and 240B can be larger or smaller than the floating gate width GA or GB to make the length of channel region CC larger or smaller.
In memory transistor 200″, mini-field oxide regions 232 between control gate 250 and source/drain regions 220A and 220B reduce the undesirable capacitive coupling between floating gates 240A and 240B and respective underlying source/drain regions 220A and 220B.
In memory transistor 200″, the voltages on floating gates 240A and 240B control conductivity through end channel regions CA″ and CB″. A significant advantage of memory transistor 200″ is that channel regions CA″ and CB″ are smaller than floating gates 240A and 240B and more importantly can be smaller than the smallest feature size achievable with the photolithography and etching processes that form floating gates 240A and 240B (e.g., less than or equal to about 0.05 μm for a 0.1-μm process). The relatively larger sizes of floating gates 240A and 240B provides higher capacitive coupling between control gate 240 and floating gates 240A and 240B, which enhances programming efficiency.
The shorter effective channel lengths under floating gates 240A and 240B in memory transistor 200″ reduce the effective resistance of each end channel CA″ or CB″. Reduced effective channel resistance under each floating gate 240A and 240B can be crucial for high performance write and read operations. For a write operation, very short channel lengths CA″ and CB″ combined with a suitably high drain and control gate voltages effectively make the end channel closest to the drain of memory transistor 200″ transparent. In particular, well-known device phenomena such as punch-through and drain-induced barrier lowering (or short channel effects) make the short channel effectively transparent regardless of the charge on the overlying floating gate 240A or 240B. This results in a higher write current and increases channel hot electron injection for faster and more efficient programming.
For a read, the short channel lengths CA″ and CB″ and sufficiently high read voltages at the drain and control gate can effectively make the channel near the drain transparent through both punch-through and/or drain-induced barrier lowering. The read current is therefore primarily dependent upon the charge stored in the floating gate adjacent the source region, which provides data integrity when reading binary, analog, or multi-bit values.
Memory transistor 200′ of FIG. 2A and memory transistor 200″ of
Operations applying appropriate voltages to the terminals of memory transistor 200 can erase, program, or read memory transistor 200. One erase operation sets the charge state of both floating gates 240A and 240B to an erased state, which in the exemplary embodiment of the invention is a low or negative threshold voltage state for channel currents in both directions. A programming operation changes the charge state of one of floating gates 240A and 240B and correspondingly the threshold voltage for channel current in a direction associated with the floating gate being nearest the source-biased region. A read operation senses the threshold voltage or the amount of channel current in the direction associated with the floating gate being read.
An erase operation can be conducted using methods similar to erase methods known for conventional floating gate transistors.
Alternatively, memory cell 200 can be erased using a source-side erase such as illustrated in FIG. 4B. For this type of erase operation, the control gate erase voltage Verase1 is at ground or a negative voltage, the p-well voltage Verase2 is grounded, and source/drain voltage Verase3 and/or Verase4 is positively biased to about 5 volts or more. This biasing generally causes hole injection due to band-to-band tunneling, which neutralizes electrons stored in the floating gate. However, band-to-band tunneling and hole injection generally requires large capacity charge pumps to drive the source current and can induce charge trapping that degrades endurance. To minimize this effect, the source/drain voltage Verase3 or Verase4 can be slowly ramped up in voltage so that the voltage of the floating gate or gates being erased correspondingly increases. The source-side erase can thus be tunnel-current limited instead of voltage limited. For the source side erase, tunnel oxide 230 can be thicker. Additionally, one or both of floating gates 240A and 240B can be erased by positively biasing the source/drain regions 220A or 220B adjacent the floating gate or gates being erased.
The source-side erase process using a tunnel-current limited biasing of source/drain region 220A or 220B and a negative biasing on control gate 250 may more easily erase a memory transistor 200 to a lower threshold voltage state (e.g., a more negative threshold voltage) than can other techniques. In particular, a control gate voltage Verase1 of about −10 volts combined with a tunnel-current limited source/drain voltage Verase3 or Verase4 of about 10 volts can erase a floating gate to achieve a negative threshold voltage. This erase technique could also be applied to conventional floating gate transistors or split gate memory cells, particularly to achieve negative threshold voltage.
With any of the above erase methods, a verify operation can determine when the erase operation has driven the threshold voltages of memory transistors to the desired erased level. One type of verify operation senses bit line current while the word lines are at a voltage corresponding to the target threshold. The direction of the current through the memory transistors can be switched during the verify operation. Bit line current below a sensing threshold for current in both directions indicates all of the floating gates have reached the target erased state.
The erase process can “over-erase” memory transistor 200 so that floating gates 240A and 240B are positively charged and the threshold voltages associated with channel regions CA and CB are near or below 0 volts. Generally, erase voltages are chosen according to the desired threshold voltage for the erased state of memory transistor 200. For a very low threshold voltage (e.g., −3.3 volts or lower), the negative gate source-side erase process uses erase voltages Verase1 and Verase3/Verase4 that are about −10 volts and +10 volts respectively to positively charge floating gates 240A and 240B. Channel regions CA and CB would then operate in depletion mode at least in the erased state. Control gate 250 modulating central channel region CC, which has a positive threshold voltage, permits low threshold voltages for data storage associated with channel regions CA and CB, without introducing unacceptable current leakage during operation of a memory array.
A memory device using conventional floating gate transistors, in contrast to memory transistors 200, must avoid erasing floating gate memory transistors to low threshold voltages. If the threshold voltage were too low in a conventional Flash memory, unselected memory cells would leak unacceptable amounts of current, especially when a large number of unselected memory cells connect to the same bit line. Accordingly, the lowest usable threshold voltage for a conventional floating gate memory device must typically be substantially greater than 0 volts, e.g., greater than about 1.5 volts. This significantly reduces the useful threshold voltage range because the upper limit of the useful threshold voltage range is often limited to no more than about 5 volts because of cell disturb and reliability limitations.
Central channel CC ideally operates as a transistor in the linear mode during the read operation. Control gate 250 directly modulates channel region CC so that the threshold voltage associated with the central channel regions CC is not subject to capacitive coupling effects between control gate 250 and an intervening floating gate. Central channel region CC can additionally have a lighter doping from channel regions CA and CB to adjust the threshold voltage of channel regions CC for optimal performance. The read operation of
The practical threshold voltage range usable for storing data generally depends on the supply voltage and cell disturb, data retention, and endurance effects at both extremes of the threshold voltage range. The desired read time is also a consideration in selecting the boundaries of the threshold voltage range. In particular, if a device uses a threshold voltage greater than the supply voltage (e.g., greater than 3.0 volts), reading the device may require a charge pump or voltage boosting circuit that drives the selected word line to a voltage Vr that is greater than the maximum threshold voltage (e.g., 5.0 volts). Charging the word line with a charge pump can slow the read speed in random access mode.
Table 1 illustrates relationships among the voltage VFG of a floating gate (e.g., 240A or 240B) when an overlying control gate 250 is at 0 volts (grounded), the type of excess charge on the floating gate, and the threshold voltage associated with the floating gate. Table 1 presumes that the channel underlying the floating gate is such that charge inversion occurs when the floating gate is at 1 volt and that control gate 250 has a 60% capacitive coupling to the floating gate.
Threshold Voltage (Word Line Voltage VWL
VWL = 0
Require to Turn On/Turn Off for +Vt/−Vt)
A conventional memory using floating gate transistors is generally limited to using a threshold voltage range starting above about 1.7 volts and extending to less than about 5.0 volts. Typically, for conventional floating-gate memory transistors, the lower limit cannot be lowered without increasing current leakage through unselected memory transistors, and the upper limit cannot be raised without sacrificing data retention and endurance, increasing cell disturb, and/or increasing the supply voltage or using a charge pump circuit to drive the selected word line. In contrast, memory transistor 200 can use a threshold voltage range from the normal upper limit (e.g., 5.0 volts) down to a lower limit that includes negative threshold voltages (e.g., −1.7 to −6.7 volts) depending on the erase scheme used. The larger threshold voltage range of memory transistor 200 facilitates storing multiple bits per floating gate because more or larger threshold voltage bins corresponding to different digital values can be fit into the larger threshold voltage range and the separation between the various threshold voltage levels is a larger percentage of the magnitude of the threshold voltage levels.
In accordance with another aspect of the invention, the threshold voltage range for data storage can be selected so that all data values correspond to floating gates 240A and 240B having positive charge (or holes). For the memory transistor of Table 1, a threshold voltage range having an upper limit of +1.7 limits the floating gates 240A and 240B to storing only positive charge or holes. Using holes for data storage can result in better data retention because the potential barrier for holes is considerably higher than the potential barrier for electrons. Lowering the upper limit of the threshold voltage range reduces cell disturb but may also reduce the punch-through effect since the channel region CB or CA underlying the unselected floating gate 240B or 240A in the read operations of
Memory transistors in this embodiment of the invention can store more information per floating gate and provide more dense storage than do conventional memories. For comparisons of the relative integrated circuit area required per bit,
Memory cell 500 of
Memory transistors 200 can store one or more bits per floating gate, which reduces the circuit area per bit of stored data. Table 2 indicates the area per bit for one, two, three, and four bits per floating gate. As indicated in Table 2, at two bits per floating gate, memory transistor 200 provides 100% and 20% more storage per circuit area than do binary memory cell 500 and 2-bit memory cell 100, and at four bits per floating gate, memory transistor 200′ provides 4 times and 2.5 times as much storage when compared to memory cells 500 and 100, respectively.
Circuit Area per Bit for a Twin-Floating-Gate Memory Transistor
Number of Bits per Floating Gate
Circuit Area per Bit
The area per bit for memory transistor 200 (
A variety of memory array architectures are available for assembling dual-floating gate memory transistors into memory arrays. One such memory array architecture is referred to herein as a contactless memory array architecture. In a contactless architecture, each memory transistor has source/drain regions that are portions of diffused lines in the substrate, and instead of having contacts from overlying layers directly to the source/drain regions of each memory transistors, banks of memory transistors generally have such electrical connections only at the ends of the banks. This can reduce the integrated circuit area per memory transistor by reducing the required contact area.
Diffused lines 620 are regions of n+ doping in p-well 620. Diffused lines 620 can be formed of salicide that is buried in semiconductor substrate 210 to reduce the resistance of diffused lines 620. Formation of salicide for diffusion regions is well known for high-speed logic and memory processes, including contactless Flash memory arrays. Alternatively, metal lines (not shown) can periodically strap diffusion lines 620 to reduce resistance. Portions of the diffused lines 620 that are under word lines 650 form source/drain regions 220 of memory transistors 200.
Each memory transistor 200 includes source/drain regions 220, a channel region 215, a pair of floating gates 240, and a control gate as described above. Channel regions 215 are in the p-well 720 and between diffused lines 620. Isolation structures such as shallow trench isolation, field oxide isolation, and/or heavily doped p+ field implant regions (not shown) separate channel regions that are in the same column of bank 600, i.e., between the same pair of diffused lines 620.
Floating gates 240 are between respective channel regions and associated word lines 650. Each floating gate 240 corresponds to a different storage location and is charged according to the value stored at that storage location. Floating gates 240 are typically formed from a first polysilicon layer, and a thin insulator layer such as a tunnel oxide layer separates the floating gates 240 from respective end regions CA and CB of channel 215 in memory transistors 200. Patterning of a second polysilicon layer forms word lines 650 that are over floating gates 240 with an insulating layer, typically an oxi-nitride-oxide (ONO) layer, between word lines 650 and the underlying floating gates 240. Word lines 650 also extend into gaps between pairs of the floating gates to modulate central portions CC of channel regions 215 as described above. The oxide separating control gate 650 and channel region CC can be formed or processed separately from the tunnel oxide under floating gates 240 to provide a greater oxide thickness similar to that found in some split-channel Flash memory cells.
As shown in
Connections of bank select cells 670 to respective pairs of diffused lines 620 are staggered relative to the connections of bank select cells 671 to respective pairs of diff-used lines 620. More specifically, bank select cell 670-0 is between metal line 690-0 and diffused line 620-0. Bank select cell 671-0 is between metal line 690-0 and a pair of diffused lines 620-0 and 620-1. Bank select cell 670-1 is between metal line 690-1 and a pair of diffused lines 610-1 and 610-2, and bank select cell 671-1 is between metal line 690-1 and a pair of diffused lines 610-2 and 610-3. This pattern continues up to bank select cell 670-N/2, which is between metal line 690-N/2 and diffused lines 620-(N-1) and 620-N, and bank select cell 671-N/2, which is between metal line 690-N/2 and diffused line 620-N.
Bank select lines 660 and 661, which can be formed from the polysilicon layer (typically poly2) forming word lines 650 or from a polysilicon layer forming peripheral transistors, respectively control bank select cells 670 and 671. Activation of a select signal BS0 on bank select line 660 simultaneously turns on all bank select cells 670 in bank 600, so that bank select cells 670 electrically connect metal lines 690 to diffused lines 620. Activation of a bank select signal BS1 on bank select line 661 simultaneously turns on all bank select cells 671 in bank 600, and bank select cells 671 electrically connect metal lines 690 to diffused lines 620.
The architecture of bank 600 and particularly the connection of metal lines 690 to diffused lines 620 in bank 600 are similar to and amenable to variations as in architectures and connections described in a co-owned U.S. Pat. No. 6,570,810, entitled “Contactless Flash Memory With Buried Diffusion Bit/Virtual Ground Lines” and co-owned U.S. Pat. No. 6,480,422, entitled “Contactless Flash Memory With Shared Buried Diffusion Bit Line Architecture”, which are hereby incorporated by reference in their entirety. Bank 600, however, uses dual-floating gate memory transistors, rather than conventional floating gate transistors as in prior contactless memory arrays. As a result, bank 600 has half as many diffused lines 620 and half as many metal lines 690 per floating gate 240 as would a similar array of conventional floating gate transistors, and if the same size floating gates are used, the pitch of metal lines 690 is about twice as wide as the pitch of similar metal lines in banks of conventional floating gate transistors. Accordingly, bank 600 advantageously provides less capacitive coupling between metal bit/virtual ground lines 690. More importantly, the wider pitch for the global metal lines 690 provides more area for layout of array supporting circuits such as column decoding and driving circuits that connect to metal lines 690.
For the access operation of
Unselected word lines 650-1 to 650-M are biased low (e.g., grounded). The low voltage shuts off all memory transistors 200 in the unselected rows because the low control gate voltage stops charge inversion in the central channel regions CC of all memory transistors 200 in the unselected rows. Current leakage in the unselected rows is thus avoided regardless of the threshold voltages associated with the floating gates 240.
Bank select circuitry (not shown) activates bank select signal BS0 on bank select line 660 to turn on bank select cells 670 in response to an address signal indicating that selected memory transistor 200 is in an even column (e.g., column 2) of bank 600. As a result, bank select cells 670 connect metal lines 690 to diffused lines 620 and particularly connects metal lines 690-1 and 690-2 to diffused lines 620-2 and 620-3, which are on opposite sides of selected memory transistor 200 in column 2. Select signal BS1 on bank select line 661 is deactivated to turn off bank select cells 671.
Column decoding and drive circuits (not shown) drive metal line 690-2 to a low voltage VL (typically ground) and drive metal line 690-1 to a higher voltage VH. Voltage VH is programming voltage Vw (typically about 4.5 to 6 volts) for a programming operation and is the read bias voltage Vbais (typically about 1 to 2 volts) for a read operation.
During the access, bank-select cells 670 applies voltage VH from metal line 690-1 to two diffused lines 620-2 and 620-1. To avoid driving current through memory transistors that are to the left of the selected memory transistor 200 in
As mentioned above, the direction of current 810 of
Accesses to memory transistors in odd columns require activating select signal BS1 and deactivating select signal BS0, so that bank select cells 671 are “on” and bank select cells 670 are “off.”
Masked 930 is stripped off or otherwise removed and a second mask 950 is formed as shown in FIG. 9B. Mask 950 defines the boundaries of floating gates 240, which overlap the implanted regions 940. Polysilicon layer 920 is etched away through openings in mask 950 to leave polysilicon floating gates 240 as shown in FIG. 9C. At this point, polysilicon for floating gates 240 can extend continuously along the direction of implanted regions 940. One or more oxidation process applied to the structure of
As noted herein, embodiments of the invention can achieve several advantages over current non-volatile memory designs. In particular, when compared to memory cells using charge trapped in insulators, memory transistors in accordance with the present invention store orders of magnitude more charge, which facilitates more precise control of threshold voltages for storage of analog or multi-bit values. When compared to conventional, floating gate transistors, embodiments of the invention permit use of a wider range of threshold voltages for data storage because central regions in the twin-floating gate transistors provide a sure shut off of unselected memory cells, which allows the floating gate to be erased to depletion (storing holes) which in turn provides a larger threshold voltage range for analog or multi-bit storage. Additionally, arrays of twin-floating gate transistors require fewer metal and diff-used bit/virtual ground lines than do similar arrays of conventional floating gate transistors and thus reduce capacitive couplings that can introduce noise and slow memory operation, and to provide a wider pitch for the layout of array supporting circuits such as column decoders and drivers.
Although the invention has been described with reference to particular embodiments, the description provides examples of the invention's application and should not be taken as a limitation. In particular, although specific memory transistors, array architectures, and access operations are described, each may be used separately from the others. The described memory transistor structures, for example, can be employed in other array structures or with alternative erase, write, or read operations. The array structure applied herein to the dual-floating gate memory transistors can also be applied to other memory transistor structures including, for example, the 2-bit memory cells of FIG. 1. Various other adaptations and combinations of features of the embodiments disclosed are within the scope of the invention as defined by the following claims.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US3895360||Jan 29, 1974||Jul 15, 1975||Westinghouse Electric Corp||Block oriented random access memory|
|US4016588||Dec 24, 1975||Apr 5, 1977||Nippon Electric Company, Ltd.||Non-volatile semiconductor memory device|
|US4173766||Sep 16, 1977||Nov 6, 1979||Fairchild Camera And Instrument Corporation||Insulated gate field-effect transistor read-only memory cell|
|US4173791||Sep 16, 1977||Nov 6, 1979||Fairchild Camera And Instrument Corporation||Insulated gate field-effect transistor read-only memory array|
|US4281397||Oct 29, 1979||Jul 28, 1981||Texas Instruments Incorporated||Virtual ground MOS EPROM or ROM matrix|
|US4360900||Sep 2, 1980||Nov 23, 1982||Texas Instruments Incorporated||Non-volatile semiconductor memory elements|
|US4380057||Oct 27, 1980||Apr 12, 1983||International Business Machines Corporation||Electrically alterable double dense memory|
|US4527257||Aug 25, 1982||Jul 2, 1985||Westinghouse Electric Corp.||Common memory gate non-volatile transistor memory|
|US4586163||Jul 1, 1983||Apr 29, 1986||Toshiba Shibaura Denki Kabushiki Kaisha||Multi-bit-per-cell read only memory circuit|
|US4630085||Feb 27, 1985||Dec 16, 1986||Nec Corporation||Erasable, programmable read-only memory device|
|US4667217||Apr 19, 1985||May 19, 1987||Ncr Corporation||Two bit vertically/horizontally integrated memory cell|
|US4742491||Sep 26, 1985||May 3, 1988||Advanced Micro Devices, Inc.||Memory cell having hot-hole injection erase mode|
|US4847808||Apr 22, 1987||Jul 11, 1989||Nec Corporation||Read only semiconductor memory having multiple bit cells|
|US4870470||Oct 16, 1987||Sep 26, 1989||International Business Machines Corporation||Non-volatile memory cell having Si rich silicon nitride charge trapping layer|
|US5021999||Dec 9, 1988||Jun 4, 1991||Mitsubishi Denki Kabushiki Kaisha||Non-volatile semiconductor memory device with facility of storing tri-level data|
|US5159570||May 7, 1991||Oct 27, 1992||Texas Instruments Incorporated||Four memory state EEPROM|
|US5172338||Apr 11, 1990||Dec 15, 1992||Sundisk Corporation||Multi-state EEprom read and write circuits and techniques|
|US5204835||Jun 13, 1990||Apr 20, 1993||Waferscale Integration Inc.||Eprom virtual ground array|
|US5214303||Jan 28, 1992||May 25, 1993||Sharp Kabushiki Kaisha||Semiconductor device ROM having an offset region|
|US5267194||Aug 30, 1991||Nov 30, 1993||Winbond Electronics Corporation||Electrically erasable programmable read-only-memory cell with side-wall floating gate|
|US5268861||Mar 3, 1992||Dec 7, 1993||Sharp Kabushiki Kaisha||Semiconductor read only memory|
|US5293563||Dec 12, 1991||Mar 8, 1994||Sharp Kabushiki Kaisha||Multi-level memory cell with increased read-out margin|
|US5295108||Apr 8, 1993||Mar 15, 1994||Nec Corporation||Electrically erasable and programmable read only memory device with simple controller for selecting operational sequences after confirmation|
|US5305262||Sep 9, 1992||Apr 19, 1994||Kawasaki Steel Corporation||Semiconductor integrated circuit|
|US5311049||Oct 13, 1992||May 10, 1994||Rohm Co., Ltd.||Non-volatile semiconductor memory with outer drain diffusion layer|
|US5315541||Jul 24, 1992||May 24, 1994||Sundisk Corporation||Segmented column memory array|
|US5345425||Nov 19, 1991||Sep 6, 1994||Fujitsu Limited||Semiconductor memory device|
|US5349221||Oct 20, 1992||Sep 20, 1994||Rohm Co., Ltd.||Semiconductor memory device and method of reading out information for the same|
|US5359554||Jun 30, 1992||Oct 25, 1994||Matsushita Electric Industrial Co., Ltd.||Semiconductor memory device having an energy gap for high speed operation|
|US5364806||Oct 12, 1993||Nov 15, 1994||Hyundai Electronics Industries Co., Ltd.||Method of making a self-aligned dual-bit split gate (DSG) flash EEPROM cell|
|US5412601||Aug 30, 1993||May 2, 1995||Nippon Steel Corporation||Non-volatile semiconductor memory device capable of storing multi-value data in each memory cell|
|US5414693||Jul 1, 1994||May 9, 1995||Hyundai Electronics Industries Co., Ltd.||Self-aligned dual-bit split gate (DSG) flash EEPROM cell|
|US5418743||Dec 6, 1993||May 23, 1995||Nippon Steel Corporation||Method of writing into non-volatile semiconductor memory|
|US5424567||May 15, 1991||Jun 13, 1995||North American Philips Corporation||Protected programmable transistor with reduced parasitic capacitances and method of fabrication|
|US5424978||Mar 14, 1994||Jun 13, 1995||Nippon Steel Corporation||Non-volatile semiconductor memory cell capable of storing more than two different data and method of using the same|
|US5426605||Aug 18, 1993||Jun 20, 1995||U.S. Philips Corporation||Semiconductor memory device|
|US5434825||Sep 3, 1993||Jul 18, 1995||Harari; Eliyahou||Flash EEPROM system cell array with more than two storage states per memory cell|
|US5477499||Oct 13, 1993||Dec 19, 1995||Advanced Micro Devices, Inc.||Memory architecture for a three volt flash EEPROM|
|US5495440||Oct 12, 1994||Feb 27, 1996||Mitsubishi Denki Kabushiki Kaisha||Semiconductor memory device having hierarchical bit line structure|
|US5518942||Feb 22, 1995||May 21, 1996||Alliance Semiconductor Corporation||Method of making flash EPROM cell having improved erase characteristics by using a tilt angle implant|
|US5521870||Dec 6, 1994||May 28, 1996||Nec Corporation||Semiconductor memory device having a coincidence detection circuit and its test method|
|US5583808||Sep 16, 1994||Dec 10, 1996||National Semiconductor Corporation||EPROM array segmented for high performance and method for controlling same|
|US5623438||Aug 26, 1996||Apr 22, 1997||Sgs-Thomson Microelectronics, Inc.||Virtual ground read only memory circuit|
|US5661053||May 25, 1994||Aug 26, 1997||Sandisk Corporation||Method of making dense flash EEPROM cell array and peripheral supporting circuits formed in deposited field oxide with the use of spacers|
|US5714412||Dec 2, 1996||Feb 3, 1998||Taiwan Semiconductor Manufacturing Company, Ltd||Multi-level, split-gate, flash memory cell and method of manufacture thereof|
|US5717636||May 3, 1996||Feb 10, 1998||Sgs-Thomson Microelectronics S.R.L.||EEPROM memory with contactless memory cells|
|US5753950||Apr 10, 1996||May 19, 1998||Motorola, Inc.||Non-volatile memory having a cell applying to multi-bit data by double layered floating gate architecture and programming/erasing/reading method for the same|
|US5760435||Apr 22, 1996||Jun 2, 1998||Chartered Semiconductor Manufacturing, Ltd.||Use of spacers as floating gates in EEPROM with doubled storage efficiency|
|US5768192||Jul 23, 1996||Jun 16, 1998||Saifun Semiconductors, Ltd.||Non-volatile semiconductor memory cell utilizing asymmetrical charge trapping|
|US5777919||Sep 13, 1996||Jul 7, 1998||Holtek Microelectronics, Inc.||Select gate enhanced high density read-only-memory device|
|US5793666||Mar 25, 1997||Aug 11, 1998||Nec Corporation||Single-chip memory system having an array of read-only memory cells|
|US5812449||May 10, 1996||Sep 22, 1998||Hyundai Electronics Industries Co., Ltd.||Flash EEPROM cell, method of manufacturing the same, method of programming and method of reading the same|
|US5821581||Jul 19, 1993||Oct 13, 1998||Texas Instruments Incorporated||Non-volatile memory cell structure and process for forming same|
|US5825686||Feb 5, 1996||Oct 20, 1998||Siemens Aktiengesellschaft||Multi-value read-only memory cell having an improved signal-to-noise ratio|
|US5841693||Apr 25, 1997||Nov 24, 1998||Nec Corporation||Non-volatile memory using field effect transistors having dual floating gates for storing two bits per cell|
|US5844839||Jul 19, 1996||Dec 1, 1998||Texas Instruments Incorporated||Programmable and convertible non-volatile memory array|
|US5862076||Jun 30, 1992||Jan 19, 1999||Waferscale Integration, Inc.||Fast EPROM array|
|US5875128||Jun 30, 1997||Feb 23, 1999||Nec Corporation||Semiconductor memory|
|US5936887||Aug 12, 1997||Aug 10, 1999||Samsung Electronics Co., Ltd.||Non-volatile memory device with NAND type cell structure|
|US5949711||Sep 24, 1997||Sep 7, 1999||Waferscale Integration, Inc.||Dual bit memory cell|
|US5963465||Dec 12, 1997||Oct 5, 1999||Saifun Semiconductors, Ltd.||Symmetric segmented memory array architecture|
|US5966603||Jun 11, 1997||Oct 12, 1999||Saifun Semiconductors Ltd.||NROM fabrication method with a periphery portion|
|US6009013||Apr 21, 1995||Dec 28, 1999||Interuniversitair Micro-Elektronica Centrum Vzw||Contactless array configuration for semiconductor memories|
|US6011725||Feb 4, 1999||Jan 4, 2000||Saifun Semiconductors, Ltd.||Two bit non-volatile electrically erasable and programmable semiconductor memory cell utilizing asymmetrical charge trapping|
|US6030871||May 5, 1998||Feb 29, 2000||Saifun Semiconductors Ltd.||Process for producing two bit ROM cell utilizing angled implant|
|US6037226||May 15, 1998||Mar 14, 2000||Lg Semicon Co., Ltd.||Method of making contactless nonvolatile semiconductor memory with asymmetrical floating gate|
|US6133098||May 17, 1999||Oct 17, 2000||Halo Lsi Design & Device Technology, Inc.||Process for making and programming and operating a dual-bit multi-level ballistic flash memory|
|US6134156||Feb 4, 1999||Oct 17, 2000||Saifun Semiconductors Ltd.||Method for initiating a retrieval procedure in virtual ground arrays|
|US6151248||Jun 30, 1999||Nov 21, 2000||Sandisk Corporation||Dual floating gate EEPROM cell array with steering gates shared by adjacent cells|
|US6169307||Dec 8, 1998||Jan 2, 2001||Matsushita Electronics Corporation||Nonvolatile semiconductor memory device comprising a memory transistor, a select transistor, and an intermediate diffusion layer|
|US6201282||Dec 23, 1999||Mar 13, 2001||Saifun Semiconductors Ltd.||Two bit ROM cell and process for producing same|
|US6215148||May 20, 1998||Apr 10, 2001||Saifun Semiconductors Ltd.||NROM cell with improved programming, erasing and cycling|
|US6243289||Apr 8, 1998||Jun 5, 2001||Micron Technology Inc.||Dual floating gate programmable read only memory cell structure and method for its fabrication and operation|
|US6248633||Oct 25, 1999||Jun 19, 2001||Halo Lsi Design & Device Technology, Inc.||Process for making and programming and operating a dual-bit multi-level ballistic MONOS memory|
|US6249454||Sep 15, 1999||Jun 19, 2001||Taiwan Semiconductor Manufacturing Company||Split-gate flash cell for virtual ground architecture|
|US6256231||Feb 4, 1999||Jul 3, 2001||Tower Semiconductor Ltd.||EEPROM array using 2-bit non-volatile memory cells and method of implementing same|
|US6261904||Jun 19, 2000||Jul 17, 2001||Advanced Micro Devices, Inc.||Dual bit isolation scheme for flash devices|
|US6266278||Aug 8, 2000||Jul 24, 2001||Sandisk Corporation||Dual floating gate EEPROM cell array with steering gates shared adjacent cells|
|US6285574||Jul 6, 1999||Sep 4, 2001||Saifun Semiconductors Ltd.||Symmetric segmented memory array architecture|
|US6317363||Oct 13, 2000||Nov 13, 2001||Sandisk Corporation||Multi-state memory|
|US6344997||May 17, 2001||Feb 5, 2002||Taiwan Semiconductor Manufacturing Company||Split-gate flash cell for virtual ground architecture|
|US6385089||May 8, 2001||May 7, 2002||Taiwan Semiconductor Manufacturing Company||Split-gate flash cell for virtual ground architecture|
|US6480422||Jun 14, 2001||Nov 12, 2002||Multi Level Memory Technology||Contactless flash memory with shared buried diffusion bit line architecture|
|US6570810||Apr 20, 2001||May 27, 2003||Multi Level Memory Technology||Contactless flash memory with buried diffusion bit/virtual ground lines|
|US20020105036||Feb 2, 2001||Aug 8, 2002||Wang Hsingya Arthur||Source side programming|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US7265409 *||Jun 14, 2005||Sep 4, 2007||Sony Corporation||Non-volatile semiconductor memory|
|US7394702 *||Apr 5, 2006||Jul 1, 2008||Spansion Llc||Methods for erasing and programming memory devices|
|US7675786 *||Mar 9, 2010||Samsung Electronics Co., Ltd.||Method of operating a semiconductor memory device having a recessed control gate electrode|
|US7834388 *||Feb 6, 2006||Nov 16, 2010||Nanostar Corporation||Memory array of non-volatile electrically alterable memory cells for storing multiple data|
|US8064267 *||Nov 14, 2008||Nov 22, 2011||Micron Technology, Inc.||Erase voltage reduction in a non-volatile memory device|
|US8391080||Oct 19, 2011||Mar 5, 2013||Micron Technology, Inc.||Erase voltage reduction in a non-volatile memory device|
|US9190162 *||Mar 13, 2012||Nov 17, 2015||Micron Technology, Inc.||Nonconsecutive sensing of multilevel memory cells|
|US20050281086 *||Jun 14, 2005||Dec 22, 2005||Toshio Kobayashi||Non-volatile semiconductor memory|
|US20060131640 *||Feb 6, 2006||Jun 22, 2006||Andy Yu||Memory array of non-volatile electrically alterable memory cells for storing multiple data|
|US20070247923 *||Apr 5, 2006||Oct 25, 2007||Meng Ding||Methods for erasing and programming memory devices|
|US20080074924 *||Nov 28, 2007||Mar 27, 2008||Andy Yu||Non-volatile electrically alterable memory cells for storing multiple data|
|US20080094917 *||Sep 14, 2007||Apr 24, 2008||Samsung Electronics Co., Ltd.||Method of operating a semiconductor memory device having a recessed control gate electrode|
|US20100124126 *||Nov 14, 2008||May 20, 2010||Vishal Sarin||Erase voltage reduction in a non-volatile memory device|
|US20110121380 *||Oct 12, 2010||May 26, 2011||Nanostar Corporation||Non-volatile electrically alterable memory cell for storing multiple data|
|US20140177335 *||Mar 13, 2012||Jun 26, 2014||Micron Technology, Inc.||Nonconsecutive sensing of multilevel memory cells|
|U.S. Classification||365/185.18, 365/185.24, 365/185.27, 257/E27.103, 257/E29.308, 257/E21.682, 257/E21.209|
|International Classification||G11C16/04, H01L29/788, H01L27/115, H01L21/28, H01L21/8247, G11C11/56|
|Cooperative Classification||G11C11/5628, H01L27/11521, H01L27/115, G11C16/0458, H01L29/7887, H01L21/28273, G11C11/5642, G11C11/5635, G11C16/0475|
|European Classification||G11C16/04M2, H01L21/28F, G11C11/56D2E, H01L29/788C, G11C16/04F4P, H01L27/115, G11C11/56D4, G11C11/56D2, H01L27/115F4|
|Jul 28, 2005||AS||Assignment|
Owner name: MULTI LEVEL MEMORY TECHNOLOGY (NEVADA), NEVADA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:MULTI LEVEL MEMORY TECHNOLOGY (CALIFORNIA);REEL/FRAME:016800/0640
Effective date: 20050720
|Aug 24, 2005||AS||Assignment|
Owner name: INTERNATIONAL DEVELOPMENT AND LICENSING, CALIFORNI
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:MULTI LEVEL MEMORY TECHNOLOGY;REEL/FRAME:017176/0517
Effective date: 20050802
|Feb 3, 2006||AS||Assignment|
Owner name: SAMSUNG ELECTRONICS CO., LTD., KOREA, REPUBLIC OF
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:INTERNATIONAL DEVELOPMENT AND LICENSING;REEL/FRAME:017230/0319
Effective date: 20050804
|Dec 5, 2008||FPAY||Fee payment|
Year of fee payment: 4
|Dec 26, 2012||FPAY||Fee payment|
Year of fee payment: 8